Power Plane Inset and the 20H Rule
Open a well-designed multilayer board and the power plane is often slightly smaller than the plane above or below it. The difference is deliberate. It exists because the field between a plane pair is not entirely contained inside the board: at the cut edge, part of it escapes into the space around the panel and radiates. Pulling the power plane back from the edge is a direct way of reducing how much of that field reaches open air.
The 20H Rule
The guidance that describes this is known as the 20H rule. The quantity H is the dielectric thickness between the power plane and the ground plane, and the rule states that the power plane’s edge should be inset from the ground plane’s edge by at least twenty times that thickness.
The relationship is geometric rather than electrical in origin. The field between two planes fringes outward at the boundary, and the distance it fringes is proportional to the separation between them. Setting the inset as a multiple of that separation therefore places the power plane boundary far enough inside the ground plane boundary that the fringing field stays under the ground plane, where it is bounded by copper rather than by air.
What Happens at an Uninset Edge
Where the two planes end at the same place, the field between them is unconstrained at the boundary. Part of the energy that should be travelling within the plane pair escapes, and the edge of the board behaves as a slot radiator fed by whatever switching currents are present between the planes.
This is not only a high-frequency effect. The plane pair carries the supply current of every fast device on the board, and that current contains components across a wide band. Where a plane pair resonates — a condition that depends on the plane dimensions and the dielectric between them — the edge is where the energy exits, and the radiation is measured in the emissions test as a set of peaks that do not obviously correspond to any component.
Insetting the power plane means that the fringing field terminates on the ground plane, which extends past it. The energy stays inside the board and dissipates in the dielectric and in the copper rather than leaving through the edge.
How Much Difference It Makes
The gains are quantified in the usual references, and the numbers explain why the rule is expressed as a multiple rather than as an absolute distance. An inset of 20H removes roughly 70 per cent of the field that would otherwise radiate from the edge. Pushing the inset to 100H raises the containment to around 98 per cent.
The rule of thumb is therefore not a threshold but a curve. Going from zero inset to 20H is the step that matters most; going beyond it keeps improving the result, with diminishing returns and a steep cost in board area.
Why It Is Often Approximated
A strict 20H inset is not always achievable. On a dense board with a dielectric of modest thickness between the planes, twenty times that thickness can be a substantial distance — large enough that removing it from the power plane leaves no room to route or to place the vias the plane needs.
The compromise most designs reach is a fixed inset of around a millimetre, which captures much of the benefit at a fraction of the area, or an inset applied only on the edges where the routing allows it.
The rule also has preconditions that are often overlooked. It applies where the power plane sits inside the board with a ground plane on each side of it, and where those ground planes extend outward by the same 20H distance. And it becomes significant on boards with enough layers for the planes to be genuinely internal, which in practice means constructions of eight layers and above. On a four-layer board, where the planes are close to the outer surfaces, the edge behaves differently and the inset has less effect.
What to Do When an Inset Is Not Possible
Where the power plane has to extend further than the ideal, the edge can still be treated deliberately.
The most direct method is to tie the ground planes together along the board edge with a line of stitching vias. Two ground planes joined at short intervals around the perimeter form a conductive boundary that contains the field inside the board, and the effectiveness of the boundary depends on the spacing between the vias relative to the frequency that has to be contained. This approach also has the practical benefit of being easier to rout than a large inset.
The second method is placement rather than structure: keep the circuits that generate the strongest and fastest currents away from the board edge, so that the field reaching the edge from those circuits is weaker in the first place.
And the third is to make the edge itself less attractive as a radiator. Where the design leaves the ground planes extending past the power plane on the outer layers as well, the field sees copper in every direction.
Related Practices at the Board Edge
Three other habits belong with the same subject.
The first is the outer layer treatment. On a design where the internal planes have been inset, the copper on the outer layers is often left extending to the board outline and connected to ground. It provides a continuous reference at the edge and gives any field that reaches that region a copper surface to terminate on rather than an open boundary.
The second is edge plating. A metalised edge connects the ground planes around the perimeter of the board, which is a stronger version of the same idea as stitching vias and is used where containment matters more than cost. It is a fabrication feature that has to be requested explicitly and it constrains the outline, so it belongs in the stack-up discussion rather than in a later amendment.
The third is the way cables leave the board. A connector at the edge of a panel carries the currents that leave the product, and those currents are a well-known path for EMI: a cable bonded to a ground structure behaves as an antenna driven by whatever potential difference exists between the two. Keeping the filtering and the ground bonding of an interface near where the cable attaches, and keeping that region away from the noisiest circuits on the board, addresses the same problem from a different direction.
Where the Decision Belongs
The inset is decided when the stack-up is arranged, because it is a property of the plane outlines rather than of the routing. It also interacts with the layer assignment, since the position of the power plane and the thickness of the dielectric on each side of it determine what the inset has to be. The considerations behind that arrangement are described in the article on layer assignment in multilayer boards.
The measurement itself belongs to the emissions testing that follows, since the edge radiation it addresses is one contributor among several to the radiated emissions of a finished product. The design-stage approach to the wider subject is covered in the article on EMC design from the layout stage, and the plane and edge items that belong in a review are listed in our layout quality checklist.
FAQ
What does the H in 20H refer to? The dielectric thickness between the power plane and the ground plane it is paired with, not the board thickness and not the layer spacing in general.
Is 20H a requirement? It is a guideline with a quantified benefit. It removes most of the edge radiation, and a smaller inset captures part of that benefit at much lower cost in board area.
When does the rule not apply? Where the planes are close to the outer surfaces, as on a four-layer board, and where there is no ground plane extending beyond the power plane to terminate the field.
